NVIDIA NeMo Guardrails
NVIDIA NeMo Guardrails is an open-source Python library for adding programmable input, dialog, retrieval and output checks to LLM applications.
What NVIDIA NeMo Guardrails does
NeMo Guardrails sits between an application and the language model it calls. Checks, which the project calls rails, run at five points: on the user's input, on the dialog flow, on chunks retrieved for RAG, on the inputs and outputs of custom actions, and on the model's output. The catalog of ready-made rails covers content safety, jailbreak detection, topic control, PII detection and masking, tool-call validation, fact-checking and LLM self-check prompts, and it can call NVIDIA's own safety models, community models or third-party APIs.
Rails are configured in YAML files and in Colang, the project's own language for describing conversational flows and guardrail logic. Custom checks are written as Python functions. An application uses it either through the Python SDK or through the bundled guardrails server, which exposes an HTTP endpoint in the style of OpenAI chat completions; there are also integrations for LangChain and LangGraph.
NVIDIA NeMo Guardrails: owner, license, hosting and price
| Fact | Value | Source and date checked |
|---|---|---|
| Owner | NVIDIA Corporation. | SOURCE ↗ |
| License | Apache 2.0. | SOURCE ↗ |
| Self-hosting | Yes, it only runs on your own infrastructure: as a Python package, or as a server from a Docker image you build from the repository. | SOURCE ↗ |
| Hosted regions | No hosted service is described. The docs list a Python SDK, a guardrails server you run, and framework integrations. | SOURCE ↗ |
| Free tier | The whole library is free to use under the Apache 2.0 license. | SOURCE ↗ |
| Pricing model | No charge for the library. The separate NeMo Guardrails microservice is described as available for the NVIDIA AI Enterprise plan. | SOURCE ↗ |
| Lowest paid plan | None for the library. The guardrails docs state no price for the NVIDIA AI Enterprise plan that covers the microservice. | SOURCE ↗ |
Each value was read from the vendor's own page on the date shown. How listings are checked
NVIDIA NeMo Guardrails pricing history
NVIDIA NeMo Guardrails has no public pricing page for this site to watch.
Where NVIDIA NeMo Guardrails fits
Teams building in Python that want to control the flow of a conversation, not only filter single messages: dialog rails and Colang flows are specific to this library. It is Apache 2.0, runs on CPU, and the same configuration works through the Python SDK or the bundled server.
Limits of NVIDIA NeMo Guardrails
It is a library to run and operate yourself: the docs describe the bundled server as self-managed with no high availability out of the box, and the Docker image has to be built from the repository. Dialog logic means learning Colang, which has two versions. It needs Python 3.10 to 3.13.
Other guardrails and audit tools
-
Amazon Bedrock Guardrails
Amazon Bedrock Guardrails is a managed AWS service that filters prompts and model responses against content, topic and sensitive-data policies.
-
Google Model Armor
Model Armor is a Google Cloud service that screens LLM prompts and responses for prompt injection, harmful content, sensitive data and malicious URLs.
-
OpenAI Guardrails
OpenAI Guardrails is an open-source Python library that wraps the OpenAI client and runs configurable checks on model inputs and outputs.
-
Presidio
Presidio is an open-source Python framework that detects and anonymizes personal data in text, images and structured data.